p9120_final_project_clustering
Qihang Wu
Import data
# --- customer ---
customer_raw <- fread("./data/marketing_campaign.csv")
alone_status <- c("Widow", "Single", "Divorced", "Alone", "Absurd")
customer <- customer_raw %>% select(-c(8, 27:28)) %>%
# for missing
mutate(Income = ifelse(is.na(Income), round(mean(Income, na.rm = TRUE)), Income),
Age = 2022 - Year_Birth,
AcceptedComp = ifelse(rowSums(across(starts_with("AcceptedCmp"))) + Response > 0, 1, 0),
Education = recode(Education, Basic = "Below some college", `2n Cycle` = "Associate degree",
Graduation = "Bachelor's degree", Master = "Master's degree", PhD = "Doctoral degree"),
Numchild = Kidhome + Teenhome,
MntTotal = MntWines + MntFruits + MntMeatProducts + MntFishProducts + MntSweetProducts + MntGoldProds,
IsAlone = ifelse(Marital_Status %in% alone_status, 1, 0),
FamilySize = ifelse(IsAlone == 1, 1, 2) + Numchild
) %>%
filter(Marital_Status != "YOLO") %>%
select(1, 27, 3, 31, 5, 29, 32, 8, 25, 9:15, 28, 16:19) %>%
mutate_at(c(3:4, 9, 17), .funs = ~as.factor(.))
# --- big five ---
# Import original data & sample 0.2%
# bf_dat <- fread("./data/data-final.csv")
# chose_ind <- sample(c(1:nrow(bf_dat)), size = round(nrow(bf_dat) * 0.002), replace = FALSE)
# bf <- bf_dat[chose_ind, ]
# write_csv(bf, "./data/big_five.csv")
big_five_raw <- fread("./data/big_five.csv")
country_df <- countrycode::codelist %>%
select(3, 6, 17)
colnames(country_df)[3] <- "country"
big_five_raw <- left_join(big_five_raw, country_df, by = "country") %>%
select(-c(51:104, 106, 108))
big_five_raw <- big_five_raw %>%
mutate(num_zero = rowSums(big_five_raw[, 1:50] == 0)) %>%
filter(num_zero == 0) %>% select(-57)
big_five <- big_five_raw %>%
filter(IPC == 1) %>%
mutate_at(c(1:50, 55:56), .funs = ~as.factor(.)) %>%
mutate(
testlapse = as.numeric(testelapse),
lat = as.numeric(lat_appx_lots_of_err),
long = as.numeric(long_appx_lots_of_err),
country = country.name.en
) %>%
select(-c(51:54, 56)) %>%
na.omit()
big_five <- big_five %>% mutate(id = seq(1:nrow(big_five))) %>%
relocate(id, .before = EXT1)EDA
# --- customer ---
# Check correlation
check_cor <- model.matrix(~., customer)[, -1]
corrplot(cor(check_cor[, -1]), method = "circle", type = "full",
tl.cex = 0.75, tl.col = "black")
summary(customer)## ID Age Education IsAlone
## Min. : 0 Min. : 26.0 Associate degree : 203 0:1444
## 1st Qu.: 2830 1st Qu.: 45.0 Bachelor's degree :1127 1: 794
## Median : 5458 Median : 52.0 Below some college: 54
## Mean : 5592 Mean : 53.2 Doctoral degree : 484
## 3rd Qu.: 8425 3rd Qu.: 63.0 Master's degree : 370
## Max. :11191 Max. :129.0
## Income Numchild FamilySize Recency Complain
## Min. : 1730 Min. :0.0000 Min. :1.000 Min. : 0.00 0:2217
## 1st Qu.: 35528 1st Qu.:0.0000 1st Qu.:2.000 1st Qu.:24.00 1: 21
## Median : 51790 Median :1.0000 Median :3.000 Median :49.00
## Mean : 52251 Mean :0.9504 Mean :2.596 Mean :49.15
## 3rd Qu.: 68307 3rd Qu.:1.0000 3rd Qu.:3.000 3rd Qu.:74.00
## Max. :666666 Max. :3.0000 Max. :5.000 Max. :99.00
## MntWines MntFruits MntMeatProducts MntFishProducts
## Min. : 0.00 Min. : 0.00 Min. : 0.0 Min. : 0.00
## 1st Qu.: 23.25 1st Qu.: 1.00 1st Qu.: 16.0 1st Qu.: 3.00
## Median : 173.00 Median : 8.00 Median : 67.0 Median : 12.00
## Mean : 303.92 Mean : 26.32 Mean : 167.1 Mean : 37.56
## 3rd Qu.: 504.75 3rd Qu.: 33.00 3rd Qu.: 232.0 3rd Qu.: 50.00
## Max. :1493.00 Max. :199.00 Max. :1725.0 Max. :259.00
## MntSweetProducts MntGoldProds NumDealsPurchases AcceptedComp
## Min. : 0.00 Min. : 0.00 Min. : 0.000 0:1630
## 1st Qu.: 1.00 1st Qu.: 9.00 1st Qu.: 1.000 1: 608
## Median : 8.00 Median : 24.00 Median : 2.000
## Mean : 27.08 Mean : 44.02 Mean : 2.323
## 3rd Qu.: 33.00 3rd Qu.: 56.00 3rd Qu.: 3.000
## Max. :263.00 Max. :362.00 Max. :15.000
## NumWebPurchases NumCatalogPurchases NumStorePurchases NumWebVisitsMonth
## Min. : 0.000 Min. : 0.000 Min. : 0.00 Min. : 0.000
## 1st Qu.: 2.000 1st Qu.: 0.000 1st Qu.: 3.00 1st Qu.: 3.000
## Median : 4.000 Median : 2.000 Median : 5.00 Median : 6.000
## Mean : 4.082 Mean : 2.664 Mean : 5.79 Mean : 5.314
## 3rd Qu.: 6.000 3rd Qu.: 4.000 3rd Qu.: 8.00 3rd Qu.: 7.000
## Max. :27.000 Max. :28.000 Max. :13.00 Max. :20.000# Total demographics
tbl_summary(dat = customer %>%
mutate(Complain = factor(Complain, levels = c(0, 1), labels = c("No", "Yes")),
IsAlone = factor(IsAlone, levels = c(0, 1), labels = c("No", "Yes")),
AcceptedComp = factor(AcceptedComp, levels = c(0, 1), labels = c("No", "Yes"))) %>%
select(-1),
label = list(Numchild ~ "Number of Children", IsAlone ~ "Is Alone", FamilySize ~ "Family Size",
Recency ~ "Number of Days Since Last Purchase", Complain ~ "Complaint (in last 2 yrs)",
# products
MntWines ~ "Wine", MntFruits ~ "Fruits", MntMeatProducts ~ "Meat Products",
MntFishProducts ~ "Fish Products", MntSweetProducts ~ "Sweet Products",
MntGoldProds ~ "Gold Products",
# campaign & place
NumDealsPurchases ~ "Number of Purchases made with a discount",
AcceptedComp ~ "Customer Accepted the Offer",
NumWebPurchases ~ "Through the Company's Website", NumCatalogPurchases ~ "Using a Catelogue",
NumStorePurchases ~ "Directly in Stores", NumWebVisitsMonth ~ "Website Visits (in last month)")) %>%
modify_table_styling(footnote = "Amount spent on a type of product in last two years",
rows = (label == "Wine"), columns = label) %>%
modify_table_styling(footnote = "Number of purchases made through a way",
rows = (label == "Through the Company's Website"),
columns = label) %>%
modify_caption("**Demographics of Customer (n=2238)**") %>%
bold_labels()| Characteristic | N = 2,2381 |
|---|---|
| Age | 52 (45, 63) |
| Education | |
| Â Â Â Â Associate degree | 203 (9.1%) |
| Â Â Â Â Bachelor's degree | 1,127 (50%) |
| Â Â Â Â Below some college | 54 (2.4%) |
| Â Â Â Â Doctoral degree | 484 (22%) |
| Â Â Â Â Master's degree | 370 (17%) |
| Is Alone | 794 (35%) |
| Income | 51,790 (35,528, 68,307) |
| Number of Children | |
| Â Â Â Â 0 | 638 (29%) |
| Â Â Â Â 1 | 1,126 (50%) |
| Â Â Â Â 2 | 421 (19%) |
| Â Â Â Â 3 | 53 (2.4%) |
| Family Size | |
| Â Â Â Â 1 | 254 (11%) |
| Â Â Â Â 2 | 762 (34%) |
| Â Â Â Â 3 | 889 (40%) |
| Â Â Â Â 4 | 301 (13%) |
| Â Â Â Â 5 | 32 (1.4%) |
| Number of Days Since Last Purchase | 49 (24, 74) |
| Complaint (in last 2 yrs) | 21 (0.9%) |
| Wine2 | 173 (23, 505) |
| Fruits | 8 (1, 33) |
| Meat Products | 67 (16, 232) |
| Fish Products | 12 (3, 50) |
| Sweet Products | 8 (1, 33) |
| Gold Products | 24 (9, 56) |
| Number of Purchases made with a discount | 2 (1, 3) |
| Customer Accepted the Offer | 608 (27%) |
| Through the Company's Website3 | 4 (2, 6) |
| Using a Catelogue | 2 (0, 4) |
| Directly in Stores | 5 (3, 8) |
| Website Visits (in last month) | 6 (3, 7) |
| 1 Median (IQR); n (%) | |
| 2 Amount spent on a type of product in last two years | |
| 3 Number of purchases made through a way | |
# --- Big Five ---
# Geographic map
mapworld <- borders("world", colour = "gray50", fill = "white")
big_five %>% ggplot() +
geom_point(aes(x = long, y = lat, color = country), size = 1.5) + mapworld +
theme_bw() + theme(legend.position = "none") +
labs(x = "Longitude", y = "Latitude", title = "Geographic Distribution of Participants (n=1191)")
# Frequency by country
bf_eda <- big_five %>%
group_by(country) %>% summarise(tot = n())
bf_eda[order(bf_eda$tot, decreasing = TRUE), ] %>%
top_n(50) %>%
ggplot(aes(x = tot, y = reorder(country, tot))) +
geom_bar(stat = "identity") + theme_bw() +
theme(legend.position = "none", axis.text.y = element_text(size = 7)) +
labs(x = "Number of Participants", y = "Country", title = "Number of Participants by Country")
Clustering on customer data
K-means
set.seed(202212)
customer_con <- customer[, c(2, 5:8, 10:16, 18:21)]
customer_scale <- scale(customer_con)
fviz_nbclust(customer_scale, FUNcluster = kmeans,
method = "gap_stat", iter.max = 50)
km <- kmeans(customer_scale, centers = 5, nstart = 20)
fviz_cluster(list(data = customer_scale, cluster = km$cluster),
ellipse.type = "convex", geom = "point",
labelsize = 5, palette = "Dark2",
main = "K-means Cluster Plot for Customer Data") + theme_bw()
K-medoids and Gower distance
# Gower distance
res_gower <- daisy(customer[, -1], metric = "gower")
summary(res_gower)## 2503203 dissimilarities, summarized :
## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 0.0000 0.1612 0.2139 0.2162 0.2689 0.5479
## Metric : mixed ; Types = I, N, N, I, I, I, I, N, I, I, I, I, I, I, I, N, I, I, I, I
## Number of objects : 2238# Find the optimal number of clusters - 3
res_silhouette <- lapply(2:10, function(x) {
pam_clust <- pam(as.matrix(res_gower), diss = TRUE, k = x)
silhouette <- pam_clust$silinfo$avg.width
})
do.call(rbind, res_silhouette) %>%
as.data.frame() %>% mutate(cluster = c(2:10)) %>%
ggplot(aes(x = cluster, y = V1)) +
geom_line() + theme_bw() +
labs(x = "Number of clusters k", y = "Sihouette Width",
title = "Optimal number of clusters")
km_mod <- pam(as.matrix(res_gower), diss = TRUE, k = 3)
# Clustering Comparison
fossil::rand.index(km$cluster, km_mod$clustering)## [1] 0.6208957# --- Demographic for each cluster ---
tbl_summary(by = clust,
data = customer %>%
mutate(clust = km_mod$clustering,
Complain = factor(Complain, levels = c(0, 1), labels = c("No", "Yes")),
IsAlone = factor(IsAlone, levels = c(0, 1), labels = c("No", "Yes")),
AcceptedComp = factor(AcceptedComp, levels = c(0, 1), labels = c("No", "Yes")),
clust = factor(clust, levels = c(1:3), labels = c("Cluster 1", "Cluster 2", "Cluster 3"))) %>%
select(-1),
label = list(Numchild ~ "Number of Children", FamilySize ~ "Family Size",
IsAlone ~ "Is Alone", Recency ~ "Number of Days Since Last Purchase",
Complain ~ "Complaint (in last 2 yrs)",
# products
MntWines ~ "Wine", MntFruits ~ "Fruits", MntMeatProducts ~ "Meat Products",
MntFishProducts ~ "Fish Products", MntSweetProducts ~ "Sweet Products",
MntGoldProds ~ "Gold Products",
# campaign & place
NumDealsPurchases ~ "Number of Purchases made with a discount",
AcceptedComp ~ "Customer Accepted the Offer",
NumWebPurchases ~ "Through the Company's Website", NumCatalogPurchases ~ "Using a Catelogue",
NumStorePurchases ~ "Directly in Stores", NumWebVisitsMonth ~ "Website Visits (in last month)")) %>%
add_p(list(all_categorical() ~ "chisq.test", all_continuous() ~ "aov")) %>% add_overall() %>%
modify_table_styling(footnote = "Amount spent on a type of product in last two years",
rows = (label == "Wine"), columns = label) %>%
modify_table_styling(footnote = "Number of purchases made through a way",
rows = (label == "Through the Company's Website"),
columns = label) %>%
modify_caption("**Demographics of Customers for each K-medoids Cluster (n=2238)**") %>%
bold_labels()| Characteristic | Overall, N = 2,2381 | Cluster 1, N = 4461 | Cluster 2, N = 6711 | Cluster 3, N = 1,1211 | p-value2 |
|---|---|---|---|---|---|
| Age | 52 (45, 63) | 53 (43, 65) | 52 (45, 62) | 51 (45, 62) | 0.2 |
| Education | 0.004 | ||||
| Â Â Â Â Associate degree | 203 (9.1%) | 38 (8.5%) | 53 (7.9%) | 112 (10.0%) | |
| Â Â Â Â Bachelor's degree | 1,127 (50%) | 222 (50%) | 353 (53%) | 552 (49%) | |
| Â Â Â Â Below some college | 54 (2.4%) | 0 (0%) | 20 (3.0%) | 34 (3.0%) | |
| Â Â Â Â Doctoral degree | 484 (22%) | 115 (26%) | 142 (21%) | 227 (20%) | |
| Â Â Â Â Master's degree | 370 (17%) | 71 (16%) | 103 (15%) | 196 (17%) | |
| Is Alone | 794 (35%) | 123 (28%) | 671 (100%) | 0 (0%) | <0.001 |
| Income | 51,790 (35,528, 68,307) | 77,040 (69,224, 82,783) | 46,734 (33,347, 61,825) | 44,155 (31,535, 57,937) | <0.001 |
| Number of Children | <0.001 | ||||
| Â Â Â Â 0 | 638 (29%) | 362 (81%) | 141 (21%) | 135 (12%) | |
| Â Â Â Â 1 | 1,126 (50%) | 74 (17%) | 368 (55%) | 684 (61%) | |
| Â Â Â Â 2 | 421 (19%) | 8 (1.8%) | 141 (21%) | 272 (24%) | |
| Â Â Â Â 3 | 53 (2.4%) | 2 (0.4%) | 21 (3.1%) | 30 (2.7%) | |
| Family Size | <0.001 | ||||
| Â Â Â Â 1 | 254 (11%) | 113 (25%) | 141 (21%) | 0 (0%) | |
| Â Â Â Â 2 | 762 (34%) | 259 (58%) | 368 (55%) | 135 (12%) | |
| Â Â Â Â 3 | 889 (40%) | 64 (14%) | 141 (21%) | 684 (61%) | |
| Â Â Â Â 4 | 301 (13%) | 8 (1.8%) | 21 (3.1%) | 272 (24%) | |
| Â Â Â Â 5 | 32 (1.4%) | 2 (0.4%) | 0 (0%) | 30 (2.7%) | |
| Number of Days Since Last Purchase | 49 (24, 74) | 42 (19, 71) | 51 (27, 75) | 50 (26, 75) | 0.002 |
| Complaint (in last 2 yrs) | 21 (0.9%) | 1 (0.2%) | 7 (1.0%) | 13 (1.2%) | 0.2 |
| Wine3 | 173 (23, 505) | 710 (464, 960) | 96 (17, 371) | 73 (15, 292) | <0.001 |
| Fruits | 8 (1, 33) | 48 (23, 93) | 6 (1, 21) | 4 (1, 15) | <0.001 |
| Meat Products | 67 (16, 232) | 427 (258, 610) | 44 (12, 142) | 31 (12, 106) | <0.001 |
| Fish Products | 12 (3, 50) | 72 (33, 130) | 8 (2, 31) | 7 (2, 21) | <0.001 |
| Sweet Products | 8 (1, 33) | 49 (24, 96) | 6 (1, 21) | 5 (1, 16) | <0.001 |
| Gold Products | 24 (9, 56) | 57 (32, 114) | 22 (8, 50) | 16 (6, 40) | <0.001 |
| Number of Purchases made with a discount | 2 (1, 3) | 1 (1, 1) | 2 (1, 3) | 2 (1, 3) | <0.001 |
| Customer Accepted the Offer | 608 (27%) | 342 (77%) | 124 (18%) | 142 (13%) | <0.001 |
| Through the Company's Website4 | 4 (2, 6) | 5 (4, 7) | 3 (2, 5) | 3 (2, 5) | <0.001 |
| Using a Catelogue | 2 (0, 4) | 6 (4, 7) | 1 (0, 3) | 1 (0, 2) | <0.001 |
| Directly in Stores | 5 (3, 8) | 8 (6, 11) | 4 (3, 7) | 4 (3, 7) | <0.001 |
| Website Visits (in last month) | 6 (3, 7) | 3 (2, 5) | 6 (4, 7) | 6 (5, 7) | <0.001 |
| 1 Median (IQR); n (%) | |||||
| 2 One-way ANOVA; Pearson's Chi-squared test | |||||
| 3 Amount spent on a type of product in last two years | |||||
| 4 Number of purchases made through a way | |||||
# --- Heatmap ---
# Products
htp_1 <- customer %>% select(10:15) %>%
mutate(cluster = km_mod$clustering) %>%
pivot_longer(
1:6, names_to = "product", values_to = "value"
) %>%
group_by(cluster, product) %>%
summarise(amount = mean(value)) %>%
ggplot(aes(x = cluster, y = product, fill = amount)) +
geom_tile() + theme_bw() +
theme(legend.position = "bottom") +
scale_y_discrete(labels = c("Fish", "Fruit", "Gold", "Meat", "Sweet", "Wine")) +
scale_fill_distiller(palette = "OrRd") +
labs(x = "Cluster", y = "Product")
# Places
htp_2 <- customer %>% select(18:20) %>%
mutate(cluster = km_mod$clustering) %>%
pivot_longer(
1:3, names_to = "place", values_to = "value"
) %>%
group_by(cluster, place) %>%
summarise(number = mean(value)) %>%
ggplot(aes(x = cluster, y = place, fill = number)) +
geom_tile() + theme_bw() +
theme(legend.position = "bottom") +
scale_y_discrete(labels = c("Catalog", "Store", "Website")) +
scale_fill_distiller(palette = "OrRd") +
labs(x = "Cluster", y = "Place")
htp_1 + htp_2 +
plot_annotation(title = "Heatmap of Amount Spend and Number of Visits by K-medoids Cluster")
# --- Networks for continuous vars: huge ---
lapply(1:length(unique(km_mod$clustering)), function(x) {
print(paste("Cluster", x, "with", sum(km$cluster == x), "subjects"))
cl <- customer_con[km_mod$clustering == x,]
net <- estimateNetwork(as.matrix(cl), default = "huge")
qgraph(net$graph, labels = names(cl), layout = "spring",
theme = "TeamFortress", label.cex = 1.5,
label.fill.horizontal = .7)
})
Clustering on big five
Latent class analysis (LCA)
# Try #class from 2 to 8
lca_dat <- big_five[, 2:51] %>% mutate_all(.funs = ~as.numeric(.))
f <- formula(paste("cbind(", paste(colnames(lca_dat), collapse = ","), ") ~ 1"))
lca_res <- mclapply(2:8, function(x) {
set.seed(202212)
lca <- poLCA(f, lca_dat, nclass = x, verbose = FALSE, nrep = 20)
return(lca)
}, mc.cores = num.cores)
df_bic <- data.frame(Class = 2:8, BIC = unlist(lapply(lca_res, function(obj) obj$bic)))
df_bic %>%
ggplot(aes(x = Class, y = BIC, label = round(BIC))) +
geom_line() + geom_point() +
geom_text(vjust = -.35) + theme_bw() +
labs(title = "BIC for Latent Class Analysis for Big Five Data")
# Choose 5 classes with smallest BIC
n_class <- df_bic$Class[which.min(df_bic$BIC)]
fit_lca <- lca_res[[n_class - 1]]Answers for each class
# --- Alluvial plots ---
# rename
score <- c("Excellent", "Very Good", "Good", "Fair", "Poor")
col_names <- c("PartyLife", "TalkLess", "SocialComfort", "Invisible", "TopicStarter",
"SayLittle", "ExtensiveContact", "AvoidAttention", "CentralPerson", "Quiet",
"Stressed", "Relaxed", "Worried", "FeelBlueLess", "Distubed",
"GetUpset", "MoodChange", "FreqMoodSwing", "Irritated", "FeelBlue",
"ConcernLess", "Interested", "Insult", "Sympathize", "NotInterestProb",
"SoftHeart", "NotInterestPpl", "TakeTime", "FeelEmo", "MakeEase",
"Prepared", "LeaveBelonging", "Meticulous", "MakeMess", "NoDelay",
"Forgetful", "LikeOrder", "ShirkDuty", "FollowSche", "Exacting",
"RichVocal", "DiffAbstract", "ViviImag", "NoInAbstract", "ExcelIdea",
"NoImag", "QuickUnderstand", "DiffWord", "SlowReflect", "FullIdea")
five_factor <- data.frame(
fact_name = c("Extraversion", "EmoStability", "Agreeableness", "Conscientious", "Openness"),
factr = c("EXT", "EST", "AGR", "CSN", "OPN"))
# set pos or neg answers
dirct <- data.frame(variable = c(paste("EXT", seq(10), sep = ""),
paste("EST", seq(10), sep = ""),
paste("AGR", seq(10), sep = ""),
paste("CSN", seq(10), sep = ""),
paste("OPN", seq(10), sep = "")),
dirct = c(rep(c(1, 0), 5),
0, 1, 0, 1, rep(0, 6),
0, 1, 0, 1, 0, 1, 0, 1, 1, 1,
1, 0, 1, 0, 1, 0, 1, 0, 1, 1,
1, 0, 1, 0, 1, 0, 1, 1, 1, 1))
# combine lca clusters
dat_clust <- lca_dat %>%
mutate(clust = fit_lca$predclass)
dat_forshow <- dat_clust %>%
pivot_longer(1:50, names_to = "variable", values_to = "value") %>%
group_by(variable, clust, value) %>%
summarise(tot = n()) %>%
pivot_wider(c(1:2), names_from = "value", values_from = "tot")
dat_forshow[is.na(dat_forshow)] <- 0
# change frequency to percentage
# convert answer to score: subtract 6 by answers of negative variables
tot_dat <- dat_forshow %>%
pivot_longer(3:7, names_to = "answer", values_to = "number") %>%
group_by(variable, answer) %>% summarise(tot = sum(number)) %>%
mutate(comb = paste(variable, answer, sep = ""))
dat_forshow_1 <- dat_forshow %>%
pivot_longer(3:7, names_to = "answer", values_to = "number") %>%
mutate(comb = paste(variable, answer, sep = ""))
dat_forshow_1 <- left_join(dat_forshow_1, dirct, by = "variable")
dat_forshow_2 <- dat_forshow_1 %>%
mutate(
answer = as.numeric(answer),
score = ifelse(dirct == 0, 6 - answer, answer)) %>%
select(1, 2, 7, 4, 5)
# change label name
dat_forshow_3 <- left_join(dat_forshow_2,
tot_dat %>% select(3, 4), by = "comb") %>%
mutate(variable = variable.x, prct = number/tot * 100) %>%
select(8, 2:3, 9) %>% pivot_wider(names_from = "score", values_from = "prct")
# change factor name
dat_forshow_4 <- dat_forshow_3 %>%
mutate(factr = substr(variable, 1, 3))
dat_forshow_5 <-
left_join(dat_forshow_4,
data.frame(name = col_names, variable = c(paste("EXT", seq(10), sep = ""),
paste("EST", seq(10), sep = ""),
paste("AGR", seq(10), sep = ""),
paste("CSN", seq(10), sep = ""),
paste("OPN", seq(10), sep = ""))),
by = "variable")
dat_forshow_fin <- left_join(dat_forshow_5, five_factor, by = "factr")
make_alluvial <- function(x){
dat_forshow_fin[, -c(1, 8)] %>%
pivot_longer(c(2:6), names_to = "Score", values_to = "Prct") %>%
mutate(clust = as.factor(clust),
Score = factor(Score, levels = c(5:1), labels = score)) %>%
filter(str_detect(fact_name, x)) %>%
ggplot(aes(y = Prct, axis1 = name, axis2 = Score, axis3 = clust)) +
geom_alluvium(aes(fill = clust), width = 1/12) +
geom_stratum(width = 1/12, fill = "black", color = "grey") +
geom_label(stat = "stratum", aes(label = after_stat(stratum))) +
scale_x_discrete(limits = c("Question", "Score", "Cluster"), expand = c(.09, .09)) +
scale_y_continuous(breaks = NULL) +
scale_fill_brewer(type = "qual", palette = "Dark2") +
guides(fill = guide_legend(title = "Cluster")) +
theme_bw() +
labs(y = "Percentage",
title = paste(x))
}
# plot alluvial
lapply(unique(dat_forshow_fin$fact_name), function(x){
make_alluvial(x)})
# combine plots
plt1 <- make_alluvial("Agreeableness")
plt2 <- make_alluvial("Conscientious")
plt3 <- make_alluvial("EmoStability")
plt4 <- make_alluvial("Extraversion")
plt5 <- make_alluvial("Openness")
ggarrange(plt1, plt2, plt3, plt4, plt5, ncol = 2, nrow = 3, common.legend = TRUE, legend = "bottom") %>%
annotate_figure(top = text_grob("Alluvial Plots"))